This research evaluates how Multimodal Large Language Models can detect and analyze joint attention—a critical element of early speech-language development—in parent-child interactions.

• Analyzed 26 parent-child interaction videos with expert annotations
• Assessed MLLMs' ability to distinguish between strong and poor joint attention
• Establishes a framework for AI-assisted assessment of developmental communication patterns
• Creates potential for earlier intervention in developmental communication disorders

This work has significant medical implications by potentially enabling broader screening for communication development issues, supporting speech-language pathologists with AI-enhanced tools, and improving early intervention for children with developmental challenges.

Towards Multimodal Large-Language Models for Parent-Child Interaction: A Focus on Joint Attention